Skip to main content

Full text of "Melting of graphene clusters"

See other formats


Melting of graphene clusters 



cn 

o 

< 
(N 



o 

I 



cd 



I 

o 
o 



> 

o 

o 
o 



X 



Sandeep K. SinghQ M. Neek-AmalQ and F.M. Peeterf^ 

Department of Physics, University of Antwerpen, 

Groenenborgerlaan 171, B-2020 Antwerpen, Belgium 

(Dated: April 23, 2013) 

Density-functional tight-binding and classical molecular dynamics simulations are used to inves- 
tigate the structural deformations and melting of planar carbon nano-clusters Cn with N=2-55. 
The minimum energy configurations for different clusters are used as starting configuration for the 
study of the temperature effects on the bond breaking/rotation in carbon lines (N<6), carbon rings 
(5<N<19) and graphene nano-fiakes. The larger the rings (graphene nano-flake) the higher the 
transition temperature (melting point) with ring-to-line (perfect-to-defective) transition structures. 
The melting point was obtained by using the bond energy, the Lindemann criteria, and the specific 
heat. We found that hydrogen-passivated graphene nano-flakes (CjvHm) have a larger melting tem- 
perature with a much smaller dependence on its size. The edges in the graphene nano-fiakes exhibit 
several different meta-stable configurations (isomers) during heating before melting occurs. 

PACS numbers: 64.70.Nd 



I. INTRODUCTION 

The study of the melting of crystals is one of the im- 
portant subjects in the field of phase transitions. Melt- 
ing phenomena occurs at the surface of bulk materials^' 
and needs a microscopic theory for a deep understanding. 
Nano-scale molecular clusters due to their size-dependent 
properties show melting processes different from those of 
bulk materials and infinite size two-dimensional materi- 
als. The melting of nano-clusters has received consid- 
erable attention recently and it was found that nano- 
clusters melt typically below their corresponding bulk 
melting temperature^iS'. This is due to the higher chem- 
ical reactivity of nano-clusters which is the consequence 
of the increased accessible surface and the presence of 
more free dangling bonds. 

The microscopic behavior of nano-clusters at finite 
temperature can be understood theoretically using a vari- 
ety of molecular dynamics (MD) methodsSti^ and can be 
determined directly by experimentpl"!^. In most of the 
simulations the microscopic structure is characterized in 
terms of bond-lengths and their average fluctuations over 
many cycles of the MD simulatiorP^Jti^. 

Since the d iscovery of two dimensional materials, i.e. 
graphenel^nS and hexagonal boron nitride sheell^ the 
melting of these new materials have attracted many re- 
searches^Sl. The new 2D crystalline materials respond to 
an increasing temperature by loosing their lattice symme- 
try, e.g. Zakharchenko et al.l^ studied the high temper- 
ature behavior of graphene using atomistic simulations. 
The melting temperature of graphene was estimated to 
be about 4900 K. Before melting first Stone- Wales defects 
appear because of their smallest energy barrier. When 
increasing temperature further eventually spaghetti-type 
of carbon chains are formed that spread in 3D. A similar 
melting process can be found for carbon nanotubes us- 
ing a much smaller critical Lindemann parameter^il. The 
melting temperature of perfect single-wall carbon nan- 
otubes (SWNTs) was estimated to be around 48001^. 



In graphene nano-ribbons different types of edges (i.e. 
zig-zag, armchair) affect the melting process differently, 
e.g. Lee et alJ22l found that at 2800 K edge recon- 
struction occurs in a zig-zag ribbon. 

In our previous work we found that the minimum en- 
ergy configuration for flat carbon clusters up to N=5 
atoms consists of a line of carbons^^l (linear chain) which 
is in agreement with ab-initicF31 calculations. Carbon pla- 
nar rings were found for 5<N<19 and graphene nano- 
fiakes are minimum energy configurations for larger N^-^. 
Here we investigate the effect of temperature on those 
minimum energy configurations and find the melting 
temperature of such small fiat carbon clusters, as func- 
tion of the size of the clusters. 

A systematic study of the size dependence of the melt- 
ing temperature is still lacking as well as the effect of 
H-passivation of the edge atoms on the melting process. 
We will present such a study and identify the different 
fundamental steps in the melting process. We found 
that graphene nano-fiakes have a lower transition tem- 
perature as compared to bulk graphene and graphene 
nanoribbons. We also found that H-passivated clus- 
ters exhibit higher melting temperature than non 
H-passivated clusters. In all cases, once clusters 
are defected they can be in different meta-stable struc- 
tures (none-planar isomers). We will compare our re- 
sults with those found for graphene and graphene nano- 
ribbons. The Lindemann index increases with re- 
spect to temperature in all cases while its slope 
versus temperature increases (decreases) linearly 
for the ring structures (graphene nano-flakes). 
Furthermore, using ab-initio molecular dynamics 
simulation we analyse the energy change due to 
defect formation. 

This paper is organized as follows. In Sec. II, we in- 
troduce the atomistic model and the simulation method. 
Sec. Ill contains our main results and a discussion of 
the melting of graphene-like clusters and H-passivated 
clusters. Sec. IV gives information on the topology of 



the defects. The effect of defects on the total energy is 
introduced in Sec. V. In Sec. VI, we conclude the paper. 



II. SIMULATION METHOD AND MODEL 
A. Minimum energy configurations 

The second-generation of Brenner reactive empirical 
bond order (REBO) potentiaP^ function between carbon 
atoms is used in the present work. All the parameters 
for the Brenner potential can be found in Ref.— and are 
therefore not listed here. 

In Fig.lllwe depict the minimum energy configurations 
for carbon clusters which are carbon lines up to 5 atoms 
(Fig. Ilja)), carbon rings for up to 18 atoms (Fig. [lib)), 
graphene nano-flakes up to 55 atoms (Fig. [ijc)) and 
hydrogen passivated graphene nano-flakes (Fig. [l](d)). 
These configurations were obtained using conjugat e gra- 
dient minimization method in our previous works^^lSSI 
The carbon line structures are energetically favorable 
structures among other possible geometries (isomers) 
which are shown in Fig. 1 of Ref.l^S, i.e. two for C3, 
6 for C4 and 11 for C5. Among all possible carbon nano- 
clusters (isomers) for 5<N<19 atoms the ground state 
are a single ring, see Fig. mb). Increasing the number 
of carbon atoms, graphene nano-flakes are formed which 
can have pentagon and heptagon defects in addition to 
common hexagons, see Fig.fTJc). Notice that by passivat- 
ing the dangling bonds by hydrogens in graphene flakes 
some structural deformations are possible. In Fig. flld) 
the minimum energy configurations for hydrogen passi- 
vated graphene nano-flakes (which were obtained by pas- 
sivating the structures in Fig. flic)) are shown. It is 
interesting to note that most of these minimum 
configuration structures have zig-zag edges which 
is due to the higher stability of these kind of edges 
as compared to arm-chair edges^-. 

In the present work we study the temperature effects 
on the structural transition and melting properties of 
these minimum energy configurations. Using molecular 
dynamic simulations, we obtain the new configuration of 
the above mentioned clusters at a given temperature T. 
This temperature is maintained during the whole simu- 
lation by the Langevin thermostapS. The MD time step 
was taken to be 0.5 fs. Different properties of the clus- 
ter were measured during the MD simulation of 10^ MD 
steps (500 ps) at fixed temperature. 



QM/MD technique based on a tight binding method 
using an approximate density-functional formalism^^^^M 
DFTB passed several benchmark tes ts with first prin- 
ciple density functional theory (DFT^pSEHMl for carbon 
structures. Alberto et al.'^'^ showed that DFTB accu- 
rately reproduced the structures and energies for a range 
of point defects such as vacancies and Stone- Wales de- 
fects in graphene. Migration barriers for vacancies and 
Stone- Wales defect formation barriers are also accurately 
reproduced. Kuc et aP^ studied the stability of 
graphene nano-flakes using DFTB and by com- 
paring their results with DFT, good agreement 
was found between the two methods. Although 
this method is two orders of magnitude faster than DFT 
but for the purpose of this work where we will study 
about 90 different configurations it will be computation- 
ally expensive. Therefore, we will use DFTB for a few 
N-values in order to show the accuracy of our classical 
MD simulation by considering one of the line carbons, 
two of the rings, five of graphene nano-flakes and six of 
H-passivated systems. 



C. Lindemann criterion and specific heat 

The root-mean-square relative bond length variance 
(Lindemann criterion) in addition to the caloric curve 
gives a reasonable computational method for determin- 
ing the melting point of nano-clusters. It is sensitive to 
any change in the bond l ength s at the microscopic scale. 
The Lindemann criteriorP^ESl is often used in molecu- 
lar dynamics and Monte Carlo simulations in order to 
estimate th e mel ting temperature in three dimensional 
bulk system426137|^ ^.^^ dimensional materials^ and nano- 
clusterpS. We used the distance-fluctuation of the Lin- 
demann index (S) in order to identify the melting tem- 
perature of our nano-clusters. For a system of N atoms, 
the local Lindemann index for the i*'* atom in the system 
is defined as"^^ ^" 



lE 



^J{rl)T-{n,)'T 






(1) 



and the system-average Lindemann index is then given 
by 



*4i:*- 



(2) 



B. Density-functional tight-binding molecular 
dynamics 

In order to have an independent test of the results ob- 
tained from the bond order potential for the melting of 
graphene nano-flakes, we also performed independent cal- 
culations using the DFTB/MD (density-functional based 
tight-binding molecular dynamics) approach which is a 



where r^- is the distance between the i and j atoms, 
N is the number of atoms and (• • •)t denotes the ther- 
mal average at temperature T. The Lindemann index HD 
depends on the specific system and its size which varies 
in the range 0.03-0.15, e.g. it was recently^S! applied to 
nanoparticles and homopolymers and found to be in the 
range of 0.03-0.05, for Ni nanoclusters it was found to be 
around 0.08^^ and for carbon nanotubes about O.OS^S. 



(a) 
(b) 

(c) 



3 4 



o o o o n 

e 7 a 9 10 

OOOOOOOO 

11 1? ia 14 IS If 17 1H 

ceo C** ■-*** 'i^-*' '••••-* *-^v I'r^ " 

19 ;!0 21 ^ 2a :^4 25 

2G 27 2R 2S 30 31 32 

'^^^ M;r XCC! iXi?^ '.=^:.' J.:..' villli 

K a 3S 35 37 3S ia 

,«XO- i*?^- X?X'^ t-ynO, X#5 I'rTi 

^ V^!:» 's:S?J 4&a ®j'; ,4'?o^ 

i.A.i^ 1,^,-,- -.i.i,' IXCJ v^/ ViiOi 

^Q 41 42 '13 44 45 

■'^'■•■■'' .'I'is". ^J^*, :*r^'i ri*:'i'i jIxISj 
1.:.:.: C:;:;::: iXOy ■^i-^:' ^^o'X tXO" 

ffi 47 4B 4y SG &1 

\ii:iXi :Ir':':':'; :':':':':': .-.'.-lOj 



''• ^ ^^ ^ X<<r v^p :;^ 

11 12 13 J£ 15 16 

% .>;<0r :^- -^ 1S^ 



^. %:? # # # 

22 2a 2i 2E. 2^ 2? 

# $> # % # 

as iA Ob HH a .as 

% 4 # 5|i ^ 



'O'^f 'OXKiJ 




FIG. 1: (Color online) The investigated configurations for carbon lines (a) carbon rings (b) and graphene nano-flakes (c). 
Pentagons (heptagons) are colored red (blue). By H-passivating the structures in (c) most of them transit to new structures 
(shown in (d)) which are indicated by underlined numbers. 



For sufficiently low temperature there is no struc- 
tural transition and the atoms exhibit thermal fluctu- 
ation around the T = equilibrium position. The os- 
cillation amplitude increases linearly with temperature 
due to Hooke's regime for the atomic vibrations leading 
to a linear increase of the Lindemann index with T. At 
higher temperature, the anharmonic vibrations (non- 
linear effects) become important and the Lindemann in- 
dex exhibits a nonlinear dependence on T. The particle 
oscillation amplitude increases faster than linear with T, 
but the system does not melt yet, since the arrangement 
of atoms have still some ordered structure, i.e. solid- 
liquid coexistence state. For small nano-clusters the lat- 
ter is related to not well defined small three dimensional 
structures. In general, melting occurs when the Linde- 
mann index increases very sharply with T over a small 
T-range. In this study we will assume that the melt- 
ing point is around the sharp jump in S, i.e. when the 
system becomes almost a random coil. We will show 
that the Lindemann index adequately indicates the 
structural deformation (melting-like transition) of car- 



bon nano-clusters. The obtained linear regime in S is 
smoother than some of the previous studiepSl for small 
nano-clusters. Therefore, we will not only use the crit- 
ical value of S to determine the melting point but we 
will also pay particular attention to the temperature de- 
pendence of the Lindemann index when identifying the 
melting temperature. 

In addition to the Lindmann index, the total energy 
(caloric curve) and specific heat variation versus T are 
two common quantities which can be used to determine 
the phase transition. We calculated the specific heat Cp 
using the equatiorP^I 



Cp{T) 



{^totallT 



(E: 



total 



kuT^ 



(3) 



where Etotai — J2i \''^i'^'i + Ep. The average potential 
energy of the system was calculated as a function of tem- 
perature. In the crystalline state the total energy of the 
system increases almost linearly with temperature, and 
then after the critical temperature is reached, it increases 
more steeply which is a signature of melting. We will 



show that for graphene nano-flake with 54 carbon atoms 
(C54 and C54H20) energy and heat capacity calculations 
are found to be consistent with the results for the melt- 
ing temperature that we obtained from the analysis of 
the Lindemann index. 



around T=4000K showing larger melting temperature 
than the corresponding non-passivated cluster which has 
a melting temperature of 3400 K (see Fig. [2]) . 



B. Lindemann index 



III. RESULTS AND DISCUSSION 



Energy 



> 

LU 




1 000 2000 3000 

Temperature (K) 



4000 



FIG. 2: (Color online) The temperature dependence of the to- 
tal energy of the graphene nanoflake C54 and the H-passivated 
C54H20 using the REBO potential. 

The temperature dependence of the total energy of 
graphene nano-flakes (C54 and C54H20) are depicted in 
Fig. [2] using the Brenner potential. For C54 the energy 
increases linearly at low temperatures and starts to de- 
viate from the linear behavior around T=2300K due to 
the reconstruction of the zigzag edges and the formation 
of pentagon-heptagon (5-7) defects. It indicates that the 
first nucleation of melting starts around 2300 K and mod- 
ifies the edges. As temperature increases, the formation 
of pentagon, heptagon, 5-7 defects or 5-8-5 are possible 
and eventually large ring structures results in a dramatic 
increase of the energy. Above T= 3400 K, there is a 
sharp increase in the energy showing a completely molten 
structure. 

For C54H20, passivation removes the dangling orbitals 
of the C atoms at the edge, lowering the reactivity, and 
increasing the stability of the cluster. It was found that 
unlike previous case (without H-passivated clusters), the 
H-passivated clusters keep their initial atomic arrange- 
ment up to higher temperature and therefore no no- 
ticeable change in the geometry was found for temper- 
atures up to T=3200 K for C54H20. The binding en- 
ergy of the H-passivated clusters are larger than for non- 
passivated clusters. As temperature increases further 
to T=3500K, some hydrogen atoms start to dissociate 
and finally the clusters convert into hydrocarbon chains 



Fig. [3] displays the variation of the Lindemann index 
with temperature for (a) C54 and (b) C54H2o. The cor- 
responding structures during heating for a few typical 
temperatures are shown in the insets. The slope of the 
function d{T) (i.e. a = ^) is plotted in Fig. li] for all 
studied structures of Fig.IlT The value of a (before reach- 
ing the melting point) increases monotonically for the 
carbon lines (Fig. flja)) and carbon rings (Fig. [lib)). 
This is an indication of keeping the initial con- 
figuration v^rhile non-linear effects indicate defect 
formation, e.g. 5-7 defects in C54. The increase of 
a is fitted in Fig. |4] by the red line for A^ < 18 which is 
given by the function a{N) = a + bx N, where: a=1.005 
(±0.032) X IQ-^K-^ and b=0.070 (±0.004) x IQ-^K''^. 
Note that increasing temperature forces the system to be 
out-of-planed, e.g. the rings at finite temperature are not 
circles and become deformed ellipsoids in 3D. 

For N > 19, there is a sudden decrease in a of size 
Aa = 1.11 X 10~^A'~^ due to the strong sp^ bonds 
within the graphene like clusters (instead of simple cova- 
lent bonds in the carbon lines and rings) and a decreasing 
number of dangling bonds. The average behavior is fitted 
by the red line a (A^) = a+6xAr, where: a= 1.63(±0.06)x 
10-^A:-i and b=-0.012 (±0.001) x lO^^AT-i. For bowl 
like clusters (N=20, 28, 38, 44), due to the presence of 
topological defective pentagon inside the cluster, the a 
value is larger as compared to their neighbor clusters. 
Therefore the important message is that dangling bonds 
and any kind of defects enhance anharmonic effects. 

In order to investigate the effect of large size sam- 
ples we also calculated a for a few large graphene nano- 
flakes and found that a decreases with N (see inset 
in Fig. ffl). The maximum considered size of graphene 
nano-flakes had 4000 atoms. For large N, one ex- 
pects saturation of a, thus a line with negative 
slope which we fitted for 19 < N < 55 should 
not be applicable. Therefore, we used the flt a{N) — 
a{oo) + a/{l + bxN), where: a(oo) = 0.134 x lO^^AT^i, 
a=1.213 (±0.005) x IQ-^K'^ and b=48.7 (±4.9) x 10^^^ 
on large clusters. These results clearly indicate that 6 for 
small graphene nano-flakes is considerably larger than for 
larger flakes and graphene. Although the Lindemann in- 
dex was defined initially in the thermodynamical limit 
(bulk material) we show that it is also a good parameter 
to investigate the effect of temperature and melting of 
nano size systems. 

For completeness, we calculated a for H-passivated 
graphene nano-fiakes for 11< N < 55 number of C- 
atoms and found that a, on average, decreases with 
N (see Fig. |4](b)). The average behavior is fitted by 
the red line a{N) = a/{l + b x N), where: a=2.262 



0.4 



X 

■D 

c 

c 
c 

CD 

E 


■D 

c 



(a)C. 










1000 2000 3000 

Temperature (K) 




1000 2000 3000 

Temperature (K) 



4000 



FIG. 3: (Color online) The temperature dependence of the Lindemann index for the cluster (a) without H-passivated C54 and 
(b) with H-passivated C54H2o- The insets show typical C (a) and C-H (b) atoms configurations where the solid areas indicate 
topological defects. 



linear ring graphene nano-flal<es 




10 20 30 40 50 

N 



H-passivated graphene nano-fial<es 



2.0 - 



^ 1 

in 

b 



1.2 - 



0.8 



(b) 

i 

] 










% 










'\ 




A 










'^-^ 


k./ 


\ A 










-^ 



10 



20 



30 



40 



50 



FIG. 4: (Color online) The low temperature rate of the Lindmann index versus the number of atoms in (a) graphene nano-flakes 
and (b) H-passivated graphene nano-flakes. The red solid lines are linear fits to the average behavior of a. The bowl-like clusters 
are shown by red solid circles (a). In the inset, the same quantity versus the number of atoms in graphene nano-flakes for N 
up to 4000 (a). 



(±0.115) X IQ-^K-^ and b=0.024 (±0.003) x 10"^ The 
rapid decrease in the latter fit (Fig. Kb)) is an indica- 
tion of the role of H-passivation in making the graphene 
nano-flakes more stable against temperature for larger N. 



C. Specific heat 



is identified as the melting temperature, showing good 
agreement with the result of the previous caloric curve 
(see Fig. [2]). A discontinuity or a sharp peak in 
the heat capacity is a clear indication of a phase 
transition. However, here this is not exactly a 
solid-to-liquid like transition, but rather a nano- 
flake to random-coil transition. 



The calculated specific heat curve for C54 is shown in 
Fig[5J A clear peak is observed in the specific heat with a 
maximum around 3400 K which we identify as the melt- 
ing temperature, and which is close to the results from 
the analysis using the Lindemann index. The specific 
heat curve is also shown for C54H20 (see red symbols 
in Fig. [5]) and displays a peak around T=4000 K which 



D. Melting-temperature 

The melting temperature for all studied clusters 
CNi2 < N < 55) is shown in Fig. ^a). The different 
parts separated by vertical dashed lines correspond to 
the three set of systems shown in Figs. fl|a-c). For small 









4000 K -T • 


12 


" 


■ c,, 


3400 K- __ T ■; 


^ — ^ 




-•-- C H 




! ■ 


(/) .„ 




64 20 




, 


*± 10 


- 






; * 


c 








' 1 


^ 








1 : 


^ 8 


- 




"1 i i 


ta 






\ ; 


L_ 














■ 


5 6 


- 






I— 






« 


CD 






f" 


^^ 4 






! /•■ ■' 


O 






2 


- 




/ 

■■ ■ 

■■■■'■■■■-■-^ •« 





•♦•♦•♦♦♦•+♦♦♦♦•♦*♦•*••**••*••*••*♦•• 



1000 2000 3000 4000 

Temperature (K) 



5000 



perature of H-passivated clusters for N=ll-55 C-atoms 
(Fig. p[b)). The melting temperature fluctuates around 
T=3500 K (note that on average T^ increases slowly with 
N with large fluctuations imposed on it) for most of the 
clusters with minima for N=29 and N=31 C-atoms due 
to the large number of defects in their structures. The 
clusters which have pentagon defects on the boundary 
usually have lower melting temperature then the others. 
Here, the melting temperature for the clusters N=13, 21, 
29, 42, 49 and 54 were also calculated using DFTB (open 
red circles in Fig. loFb)). In most cases the DFTB results 
are close to the Brenner potential results indicating that 
the Brenner bond order potential is a useful specialized 
potential for thermal effects in hydrocarbons. 



FIG. 5: (Color online) The temperature dependence of the 
specific heat for C54 and C54H2o. 



TABLE I: Melting temperature for large clusters 



iV 



T^{K) 



98 3800 

142 4000 

194 4100 

322 4200 

1000 4400 

graphene 5500 



linear structures except for the carbon dimer the melting 
temperature increases with the number of atoms (linear 
chains have bond breaking). 

In order to check if for large N we approach the melt- 
ing temperature of bulk graphene, we also calculated the 
melting temperature of three large graphene nano-fiakes 
and graphene (by performing a simulation having 4000 
atoms with periodic boundary condition with NPT en- 
semble) and presented the results in Table I. From Table 
I it is clear that large flakes approach slowly the melting 
temperature of graphene, i.e. 5500 K. As a compari- 
son the melting temperature reported in Ref.'^ by using 
the LCBOPII potential was 4900 K and in Ref.lSl using 
REBO was 5200 K. 

We fitted the melting temperature for graphene 
like clusters (red curve in Fig. [61(a)) by the func- 
tion r„(iV) = T,^,""=-(a/(fe+A^)) where a = 417(±47)x 
lO^A', h = 119.34(±17.71) and r,^f"= = 5500ii: for 
graphene was taken from our simulation. We in- 
cluded the results of Table I in this fit. We also 
calculated the melting temperature using DFTB 
for the clusters with N=5, 10, 18, 20, 28, 38, 45, 
and 54 (C54) which are represented by the open 
red circles in Fig. [61(a). They are found to slightly 
overestimate the melting temperature but exhibit 
clearly the same N-dependent trend. 

For completeness, we also calculated the melting tem- 



IV. TOPOLOGY OF DEFECTS 

In this section we consider the topology of some of the 
defects which are created during the melting process of 
C54 using DFTB calculations. We found that these de- 
fects have a pronounced effect on the melting behavior 
of the system whose mechanism is different from both 
graphene and graphene nanoribbons. At low tempera- 
ture the probability of defect creation is small and the 
flakes remains perfect which for this low temperature 
range is similar to graphene^QJ and graphene nanorib- 
borPSl. Increasing temperature (above about 2300 K) 
the energy increases and the system can overpass cer- 
tain potential barriers and we found that during the 
molecular dynamics simulation the system transits from 
one metastable state to another, see Figs. JTJJa-b). We 
show five snap shots of C54 at different temperature in 
Fig. [7] The transition temperature at which the edge 
reconstruction occurs, i.e. T^ 2400 K is lower than 
those found for defect formation in graphene, i.e. 3800 K 
and edge reconstruction in graphene nano-ribbons^^lj i.e. 
2800 K. As seen in Fig. ^^) the first defected struc- 
ture has one heptagon and one pentagon at the edge 
with different bond lengths. The energy difference be- 
tween non-defected (Fig. IWa)) and defected (Fig. [Wb)) 
one is about 0.28 eV/atom which is due to the 900 K 
difference in the temperature between these two struc- 
tures. In Fig. [7] the next defected configuration due to a 
1200 K increase in temperature is shown which has more 
heptagon and pentagon defects (colored parts). By in- 
creasing temperature further we found some transition 
in the defected parts and even an edge reconstruction 
from heptagon to hexagon and vise versa until the ap- 
pearance of a tail- like part in the cluster, FigJjFe). The 
melting temperature for C54 as shown in Fig. pTa) (open 
circle symbol) is around 3900 K. This melting tempera- 
ture is lower than those found for graphene. The larger 
the graphene nano-flake the higher the melting temper- 
ature. For small graphene nano-flakes the larger number 
of dangling bonds results in a lower melting temperature 
and larger boundary effects. The energy diagram for five 
snap shots is depicted in Fig. JT^f ) . The presented energy 



6000 
5500 
5000 
^ 4500 
"'e 4000 
3500 
3000 
2500 
2000 



linear ring 



raphene nano-flakes 



■— REBO 
■= DFTB 



(a) 




10 20 30 40 50 

N 







H-passivated graphene nano-flakes 




■- 


-REBO 


4000 


a 


DFTB 




■ 


3600 


"fl 


/ ■ ■ ... 




■ 


3200 


-■ 




1 1 






■ 


2800 


"(b) 


■ 



10 20 30 40 50 60 

N 



FIG. 6: (Color online) The melting temperature versus the number of atoms using Brenner (square symbols) and DFTB (open 
circles) for (a) nano-graphene and (b) hydrogenated nano-graphene. The error bar is 50 K. 



is the difference between the total energy of the system at 
given temperature and the zero temperature total energy 
for C54. 

In Figs. Isja-d) we show the temperature effect on 
C54H20. In Fig. Isle) the corresponding energy diagram is 
shown. Hydrogens become released at T=3900K. Notice 
that DFTB calculations for the shown configurations in 
Figs. [sFa-d) were performed separately, i.e. we do not 
increase temperature of sample Fig. ^a) in order to ob- 
tain Fig. ^b), but instead we performed four different 
calculations for these four snap shots. Thus we do not 
expect that we have sequential configurations. 



which can then be applied to understand the growth and 
thermal treatment of nanographene. We supplemented 
our analysis by DFTB calculations which confirm the N- 
dependence of the melting temperature. 



VI. ACKNOWLEDGMENTS 

This work was supported by the EU-Marie Curie 
IIF postdoc Fellowship/299855 (for M.N.-A.), the ESF- 
EuroGRAPHENE project CONGRAN, the Flemish Sci- 
ence Foundation (FWO-Vl), and the Methusalem Foun- 
dation of the Flemish Government. 



CONCLUSION 



Using molecular dynamics simulation and the Linde- 
mann index for melting supplemented with results for 
the total energy and the specific heat, we investigated 
the melting of carbon nano clusters. The melting tem- 
perature of small carbon flakes is lower than those for 
graphene and graphene nano-ribbons. The Lindemann 
index is sensitive to temperature and is a good quan- 
tity for determining when structural deformations of 
the clusters start to occur. The melting temperature 
of small flakes on average increases versus the number 
of atoms in carbon nano clusters. All clusters investi- 
gated show premelting behavior with different premelt- 
ing intervals. For certain N-values defects are already 
present inside the cluster which lowers the melting tem- 
perature. H-passivated clusters have a higher melting 
temperature than the non H-passivated clusters with the 
same number of C-atoms. The melting temperature 
for H-passivated clusters is larger than for non- 
passivated clusters. Our simulation results also help 
to understand the formation of defects (due to the in- 
crease of temperature) in the graphene nano-flake 





(ij T=15<X)K 



(bl T=24O0 K 





(c) T=270OK 



|d) T=38DOIC 




U 

1.0 

O.B 
I 

0.2 



■((] 



(a) 



(e) 





:) 


<d) 




(b) _j! 








IKi 


1.1 

O.ST 



0.^ 



0.60 



A J. 



(e}T = 39O0K 



FIG. 7: (Color online) Five different snap shots of the cluster C54 at different temperature (a-e). Bond lengths are indicated 
in the figure and the colored areas indicate defect structures, (f) The corresponding energy diagram for the five snap shots. 



* Electronic address 



* Electronic address 
^ Electronic address 



SandeepKumar. SinghQua.ac.bel 



neekamal@srttu.edu I 

Francois. PeetersQua. ac.bel 
^ J. G. Dash, Rev. Mod. Phys. 71, 1737 (1999). 
^ K. Koga, T. Ikeshoji, and K. I. Sugawara, Phys. Rev. Lett. 

92, 115507 (2004). 
^ F. Ercolessi, W. Andreoni, and E. Tosatti, Phys. Rev. Lett. 



66, 911 (1991). 
* L. J. Lewis, P. Jensen and J. L. Barrat, Phys. Rev. B 56, 

2248 (1997). 
^ G. L. Gleveland, W. D. Luedtke, and U. Landman, Phys. 

Rev. B 60, 5065 (1999). 
^ A. Rytkonen, H. Hakkinen, and M. Manninen, Phys. Rev. 

Lett. 80, 3940 (1998). 




09 • 




(a) T=3300 K 



(b) T=39(X1K 




(cl T=4D00 K 



[d|T=42O0K 



1.5 



E 1.0 



LU 



0.5 



0.0 



(?) 



0.76 



(b) 



(=) 



1.15 



1.32 



FIG. 8: (Color online) Four different snap shots of the cluster C54H20 at different temperature (a-d). Bond lengths are 
indicated in the figure and the colored areas indicate defect structures, (e) The corresponding energy diagram for the four snap 
shots. 



'' K. Joshi, D. G. Kanhere, and S. A. Blundell, Phys. Rev. 

B 66, 155329 (2002). 
^ F. C. Chuang, C. Z. Wang, S. Ogiit, J. R. Chelikowsky, 

and K. M. Ho, Phys. Rev. B 69, 165408 (2004). 
s T. L. Beck and R. S. Berry, J. Chem. Phys. 88, 3910 

(1988). 
^" B. S. Bas, M. J. Ford, and M. B. Cortie, J. Phys.: Condens. 

Matter 18, 55 (2006). 
^^ M. Schmidt, R. Kusche, W. Kronmiiller, B. von Issendorff 



and H. Haberland, Phys. Rev. Lett. 79, 99 (1997). 
^^ A. A. Shvartsburg and M. F. Jarrold, Phys. Rev. Lett. 85, 

2530 (2000). 
^^ G. A. Breaux, R. C. Benirschke, T. Sugai, B. S. Kinnear, 

and M. F. Jarrold, Phys. Rev. Lett. 91, 215508 (2003). 
^'^ G. Rey, L. J. Gallego, J. Garcia-Rodeja, J. A. Alonso, and 

M. P. Miguez, Phys. Rev. B 48, 8253 (1993). 
^^ J. Garcia-Rodeja, C. Rey, L. J. Gallego and J. A. Alonso, 

Phys. Rev. B 49, 8495 (1994). 



10 



" T. X. Li, Y. L. Ji, S. W. Yu and G. H. Wang, Solid State 

Commun. 116, 547 (2000). 
^■^ A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. 

Novoselov, and A. K. Geim, Rev. Mod. Phys. 81, 109 

(2009). 
^* A. K. Geim and K. S. Novoselov, Nat. Mater. 6, 183 (2007). 
^^ N. Alem, R. Erni, C. Kisielowski, M. D. Rossell, W. Gan- 
nett, and A. Zettl, Phys. Rev. B 80, 155425 (2009). 
^° K. V. Zakharchenko, A. Fasolino, J. H. Los, and M. I. 

Katsnelson, J. Phys.: Condens. Matter 23, 202202 (2011). 
^^ K. Zhang, G. M. Stocks, and J. Zhong, Nanotechnology 

18, 285703 (2007). 
22 G. D. Lee, C. Z. Wang, E. Yoon, N. M. Hwang, and K. M. 

Ho, Phys. Rev. B 81, 195419 (2010). 
2^ D. P. Kosimov, A. A. Dzhurakhalov, and F. M. Peeters, 

Phys. Rev. B 78, 235433 (2008). 
2^ K. Raghavachari and J. S. Binkley, J. Chem. Phys. 87, 

2191 (1987). 
2^ D. P. Kosimov, A. A. Dzhurakhalov, and F. M. Peeters, 

Phys. Rev. B 81, 195414 (2010). 
2® D. W. Brenner, O. A. Shenderova, J. A. Harrison, S. J. 

Stuart, B. Ni, and S. B. Sinnott, J. Phys.; Condens. Matter 

14, 783 (2002). 
^'^ A. Kuc, T. Heine, and G. Seifert, Phys. Rev. B 81, 085430 

(2010). 
2* L. Rui, H. Yuan-Zhong, W. Hui, and Z. Yu-Jun, Chin. 

Phys. 17, 4253 (2008). 
2^ D. Porezag, T. Frauenheim, T. Kohler, G. Seifert, and R. 

Kaschner, Phys. Rev. B 51, 12947 (1995). 
^° M. Elstner, D. Porezag, G. Jungnickel, J. Eisner, M. 

Haugk, T. Frauenheim, S. Suhai, G. Seifert, Phys. Rev. 

B 58, 7260 (1998). 
^^ T. Frauenheim, G. Seifert, M. Elstner, T. Niehaus, C. 

Kohler, M. Sternberg, Z. Hajnal, A. Di Carlo, S. Suhai, 



J. Phy.: Condens. Matter 14, 3015 (2002). 
^2 G. Zheng, S. Trie, K. Morokuma, Chem. Phys. Lett. 412, 

210 (2005). 
^^ A. Zobelli, V. Ivanovskaya, P. Wagner, L Suarez-Martinez, 

A. Yaya, and C. P. Ewels, Phys. Status Solidi B 249, 276 

(2012). 
^^ N. H. March and M. P. Tosi, Introduction to Liquid State 

Physics (World Scientific, Singapore, 2002). 
^^ J. M. Ziman, Principles of the Theory of Solids (Cambridge 

University, Cambridge 1972). 
^^ K. Sokolowski-Tinten, C. Blome, J. Blums, A. Cavalleri, 

C. Dietrich, A. Tarasevitch, L Uschmann, E. Forster, M. 

Kammler, M. Horn-von-Hoegen, and D. von der Linde, 

Nature (London) 422, 287 (2003). 
^'^ G. Shen, V. B. Prakapenka, M. L. Rivers, and S. R. Sutton, 

Phys. Rev. Lett. 92, 185701 (2004). 
3* E. C. Neyts and A. Bogaerts, J. Phys. Chem. C 113, 2771 

(2009). 
^^ Y. Zhou, M. Karplus, K. D. Ball, and R. S. Berry J. Chem. 

Phys. 116, 2323 (2002). 
■^^ F. Ding, K. Bohon, and A. Rosen, Eur. Phys. J. D 34, 275 

(2005). 
■'^ F. A. Lindemann, Physik. Z. 11, 609 (1910). H. Lowen, 

Phys. Rep. 237, 249 (1994). J. H. Bilgram, Phys. Rep. 

153, 1 (1987). 
''2 P. J. Hsu, J. S. Luo, S. K. Lai, J. F. Wax, and J. -L. 

Bretonnet, J. Chem. Phys. 129, 194302 (2008). 
*^ M. Neek-Amal and F. M. Peeters, Phys. Rev. B 83, 235437 

(2011). 
'*'' B. Steele, R. Perriot, V. Zhakhovsky, and L Oleynik, 

http://meetings.aps.Org/link/BAPS.2012.MAR.W12.3 
"^^ P. Koskinen, S. Malola, and H. Hakkinen, Phys. Rev. Lett. 

101, 115502 (2008).